Simple and Effective Dynamic Model Identification Procedure for Magnetically Suspended Flexible Rotor Systems

Advanced controllers for magnetically supported flexible rotors require an accurate model of the system dynamics. Standard identification procedures can generate accurate models, but in general, do not provide the specific value of physical parameters, needed in some applications such as inertial centering. Moreover, some of these identification techniques are not trivial for implementation and may require special signal-processing skills, nonlinear optimization, etc. In the robotics community, a well-established and widely recognized identification procedure can be found for fully-actuated industrial robots. This paper suggests a procedure that makes this identification technique applicable, also to magnetically supported flexible rotors, which are under-actuated systems. The original maximum-likelihood based procedure is modified by introducing both statistics and weights to the covariance matrix of input forces. The procedure is exemplified through numerical simulations, considering an 8-DoF flexible rotor system. The results show that although the system is under-actuated, the modified identification procedure provides satisfactory results, in the sense that all rotor parameters are accurately identified, including the parameters of mass imbalance.